---
title: "Security and trust | Agrim Intelligence"
description: "Who can do what in Omega and Agrim BOS, where your data goes, and what is planned. Written from the products' own security documents."
source: https://agrim.ai/security
---

Security and trust

# Where your data goes, and who can do what.

The questions a security review asks about Omega and Agrim BOS, answered from the products' own security documents.

1

## Where does Omega run, and where does my data go?

In your own environment.

Your environment. Omega is installed in your own environment, and the tests it writes run there.

Nothing back to us. Omega sends no usage data back to Agrim.

Every address published. Every outside address an installation may contact is listed, and a built-in check tests each connection.

Offline installation. For machines that cannot reach an image registry, Omega can be installed from a signed bundle. It still needs to reach the AI model you choose.

2

## Which AI models does Omega use, and with whose keys?

Yours.

Omega uses your own AI model keys. You can send all of its AI traffic through your own model gateway.

The model sees what the work needs: the parts of your tickets, documents and pages that a test case is written from. It goes to the model endpoint you configure.

3

## Who can do what?

Roles your admins set.

Omega: five roles. From Viewer to Super Admin. People sign in with Google, limited to your company's own domains, or with an email and password when invited.

Omega: AI tools stay out of access. The AI tools connected to Omega cannot manage people or change roles.

Agrim BOS: two roles and labels. Admins set who is an admin and who is a member, and which labels each member holds. An AI model files each meeting under those labels.

4

## Where do people decide?

At the steps that matter.

Omega: test cases. You can require a person to approve each proposed test case before it joins the suite. Tests arrive as changes in your own repository, merged under your own rules.

Omega: automated jobs. Scheduled jobs and other automated credentials cannot pass approval steps meant for a person.

Agrim BOS: anything published. Posts, comments and follows wait for an admin to approve them, and are checked for confidential details before they go out. The record keeps who approved or rejected each draft.

5

## How are secrets and releases handled?

Kept out of the database, and signed.

Stored credentials. Encrypted, with the key kept outside the database and its backups.

Your secret manager. Every secret Omega needs can be supplied from your own secret manager.

Test runs. A generated test can reach only the application under test and the addresses you list.

Releases. Every release is signed and comes with a list of its components, so you can check what you run. Dependencies are scanned for known vulnerabilities every week.

6

## Report a security issue

Write to us directly.

Write to [security@agrim.ai](mailto:security@agrim.ai). We acknowledge a report within five business days, and give a first assessment within ten business days after that.

[Ask us about anything not covered here](https://agrim.ai/contact/conversation).

## Questions from your security team? Book a technical conversation with the people who build Omega and Agrim BOS.

[Book a technical conversation](https://agrim.ai/contact/conversation) [Send us a message](https://agrim.ai/contact)
